I'm in the middle of creating a meta data management system (including
merging and persistent identifier management) for a somewhat different
domain (intranets and business integration), but it's based on Topic Maps
and so is well suited to other means of meta data handling / mangling. It's
also going to be open-source, and it might be well-suited to library tasks
as well.

So in order to test the integrity and performance of my system so far I'm
wondering if there's a suitable open dataset of bibliographic records that
aren't too obscure (meaning, I can find the titles at amazon or Open
Library) that you could recommend? More than 1000 records, but less than a
million, maybe?